ModuleNotFoundError: No module named 'examples.simultaneous'

[ereday]

🐛 Bug

Hi,

I was trying to evaluate the pre-trained models under "Efficient Wait-k Models for Simultaneous Machine Translation". For this, I followed the instructions given in the readme. Specifically, I did followings:

After downloading model and data and placing them under pre_saved:

cd ~/attn2d/pre_saved
tar xzf iwslt14_de_en.tar.gz
tar xzf tf_waitk_model.tar.gz

k=5 # Evaluation time k
output=wait$k.log
CUDA_VISIBLE_DEVICES=0 python generate.py pre_saved/iwslt14_deen_bpe10k_binaries/ -s de -t en --gen-subset test --path pre_saved/tf_waitk_model.tar.gz --task waitk_translation --eval-waitk $k --model-overrides "{'max_source_positions': 1024, 'max_target_positions': 1024}" --left-pad-source False --user-dir examples/waitk --no-progress-bar --max-tokens 8000 --remove-bpe --beam 1 2>&1 | tee -a $output

It generates following error message:

Traceback (most recent call last):
  File "generate.py", line 11, in <module>
    cli_main()
  File "/home/attn2d/fairseq_cli/generate.py", line 276, in cli_main
    parser = options.get_generation_parser()
  File "/home/attn2d/fairseq/options.py", line 33, in get_generation_parser
    parser = get_parser("Generation", default_task)
  File "/home/attn2d/fairseq/options.py", line 197, in get_parser
    utils.import_user_module(usr_args)
  File "/home/attn2d/fairseq/utils.py", line 350, in import_user_module
    importlib.import_module(module_name)
  File "/home/anaconda3/envs/py37/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1006, in _gcd_import
  File "<frozen importlib._bootstrap>", line 983, in _find_and_load
  File "<frozen importlib._bootstrap>", line 967, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 677, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 728, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/home/attn2d/examples/waitk/__init__.py", line 1, in <module>
    from . import models, tasks
  File "/home/attn2d/examples/waitk/models/__init__.py", line 7, in <module>
    importlib.import_module('examples.simultaneous.models.' + model_name)
  File "/home/anaconda3/envs/py37/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ModuleNotFoundError: No module named 'examples.simultaneous'

EDIT

Okay, here is more detail about this:

I believe thisline is responsible for the error message shared above.

I changed this importlib.import_module('examples.simultaneous.models.' + model_name) to importlib.import_module('examples.waitk.models.' + model_name)

Then, I got another error:

  File "generate.py", line 11, in <module>
    cli_main()
  File "/home/attn2d/fairseq_cli/generate.py", line 276, in cli_main
    parser = options.get_generation_parser()
  File "/home/attn2d/fairseq/options.py", line 33, in get_generation_parser
    parser = get_parser("Generation", default_task)
  File "/home/attn2d/fairseq/options.py", line 197, in get_parser
    utils.import_user_module(usr_args)
  File "/home/attn2d/fairseq/utils.py", line 350, in import_user_module
    importlib.import_module(module_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1006, in _gcd_import
  File "<frozen importlib._bootstrap>", line 983, in _find_and_load
  File "<frozen importlib._bootstrap>", line 967, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 677, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 728, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/home/attn2d/examples/waitk/__init__.py", line 1, in <module>
    from . import models, tasks
  File "/home/attn2d/examples/waitk/models/__init__.py", line 8, in <module>
    importlib.import_module('examples.waitk.models.' + model_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/home/attn2d/examples/waitk/__init__.py", line 1, in <module>
    from . import models, tasks
  File "/home/attn2d/examples/waitk/models/__init__.py", line 8, in <module>
    importlib.import_module('examples.waitk.models.' + model_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/home/attn2d/examples/waitk/models/waitk_transformer.py", line 24, in <module>
    from examples.simultaneous.modules import TransformerEncoderLayer, TransformerDecoderLayer

So, I changed this line here to ```from examples.waitk.modules import TransformerEncoderLayer, ```` too. Then when I tried once more, I got the following error:

  File "generate.py", line 11, in <module>
    cli_main()
  File "/home/attn2d/fairseq_cli/generate.py", line 276, in cli_main
    parser = options.get_generation_parser()
  File "/home/attn2d/fairseq/options.py", line 33, in get_generation_parser
    parser = get_parser("Generation", default_task)
  File "/home/attn2d/fairseq/options.py", line 197, in get_parser
    utils.import_user_module(usr_args)
  File "/home/attn2d/fairseq/utils.py", line 350, in import_user_module
    importlib.import_module(module_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1006, in _gcd_import
  File "<frozen importlib._bootstrap>", line 983, in _find_and_load
  File "<frozen importlib._bootstrap>", line 967, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 677, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 728, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/home/attn2d/examples/waitk/__init__.py", line 1, in <module>
    from . import models, tasks
  File "/home/attn2d/examples/waitk/models/__init__.py", line 8, in <module>
    importlib.import_module('examples.waitk.models.' + model_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/home/attn2d/examples/waitk/__init__.py", line 1, in <module>
    from . import models, tasks
  File "/home/attn2d/examples/waitk/models/__init__.py", line 8, in <module>
    importlib.import_module('examples.waitk.models.' + model_name)
  File "/home/anaconda3/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/home/attn2d/examples/waitk/models/waitk_transformer.py", line 25, in <module>
    from examples.waitk.modules import TransformerEncoderLayer, TransformerDecoderLayer
  File "/home/attn2d/examples/waitk/modules/__init__.py", line 2, in <module>
    from .controller import Controller

So, to fix it I commented out following lines in examples/waitk/modules/init.py:

from .controller import Controller
from .branch_controller import BranchController
from .oracle import SimulTransOracleDP, SimulTransOracleDP1

Next, I've tried to use the generation command given in the readme once more..

CUDA_VISIBLE_DEVICES=0 python generate.py pretrained-sources/iwslt14_deen_bpe10k_binaries/ -s de -t en --gen-subset test --path pretrained-sources/model.pt --task waitk_translation --eval-waitk $k --model-overrides "{'max_source_positions': 1024, 'max_target_positions': 1024}" --left-pad-source False --user-dir examples/waitk --no-progress-bar --max-tokens 8000 --remove-bpe --beam 1 2>&1 | tee -a $output

I got this error:

2021-09-20 20:29:46 | INFO | fairseq_cli.generate | Namespace(all_gather_list_size=16384, beam=1, bpe=None, checkpoint_suffix='', cpu=False, criterion='cross_entropy', data='pretrained-sources/iwslt14_deen_bpe10k_binaries/', data_buffer_size=0, dataset_impl=None, decoding_format=None, diverse_beam_groups=-1, diverse_beam_strength=0.5, diversity_rate=-1.0, empty_cache_freq=0, eval_bleu=False, eval_bleu_args=None, eval_bleu_detok='space', eval_bleu_detok_args=None, eval_bleu_print_samples=False, eval_bleu_remove_bpe=None, eval_tokenized_bleu=False, eval_waitk=5, force_anneal=None, fp16=False, fp16_init_scale=128, fp16_no_flatten_grads=False, fp16_scale_tolerance=0.0, fp16_scale_window=None, gen_subset='test', iter_decode_eos_penalty=0.0, iter_decode_force_max_iter=False, iter_decode_max_iter=10, iter_decode_with_beam=1, iter_decode_with_external_reranker=False, left_pad_source='False', left_pad_target='False', lenpen=1, load_alignments=False, log_format=None, log_interval=100, lr_scheduler='fixed', lr_shrink=0.1, match_source_len=False, max_len_a=0, max_len_b=200, max_sentences=None, max_source_positions=1024, max_target_positions=1024, max_tokens=8000, memory_efficient_fp16=False, min_len=1, min_loss_scale=0.0001, model_overrides="{'max_source_positions': 1024, 'max_target_positions': 1024}", model_parallel_size=1, momentum=0.99, nbest=1, no_beamable_mm=False, no_early_stop=False, no_progress_bar=True, no_repeat_ngram_size=0, num_shards=1, num_workers=1, optimizer='nag', path='pretrained-sources/model.pt', prefix_size=0, print_alignment=False, print_step=False, quantization_config_path=None, quiet=False, remove_bpe='@@ ', replace_unk=None, required_batch_size_multiple=8, results_path=None, retain_iter_history=False, sacrebleu=False, sampling=False, sampling_topk=-1, sampling_topp=-1.0, score_reference=False, seed=1, shard_id=0, skip_invalid_size_inputs_valid_test=False, source_lang='de', target_lang='en', task='waitk_translation', temperature=1.0, tensorboard_logdir='', threshold_loss_scale=None, tokenizer=None, truncate_source=False, unkpen=0, unnormalized=False, upsample_primary=1, user_dir='examples/waitk', warmup_updates=0, weight_decay=0.0)
2021-09-20 20:29:46 | INFO | fairseq.tasks.translation | [de] dictionary: 8848 types
2021-09-20 20:29:46 | INFO | fairseq.tasks.translation | [en] dictionary: 6632 types
2021-09-20 20:29:46 | INFO | fairseq.data.data_utils | loaded 6750 examples from: pretrained-sources/iwslt14_deen_bpe10k_binaries/test.de-en.de
2021-09-20 20:29:46 | INFO | fairseq.data.data_utils | loaded 6750 examples from: pretrained-sources/iwslt14_deen_bpe10k_binaries/test.de-en.en
2021-09-20 20:29:46 | INFO | fairseq.tasks.translation | pretrained-sources/iwslt14_deen_bpe10k_binaries/ test de-en 6750 examples
2021-09-20 20:29:46 | INFO | fairseq_cli.generate | loading model(s) from pretrained-sources/model.pt
Traceback (most recent call last):
  File "generate.py", line 11, in <module>
    cli_main()
  File "/home/attn2d/fairseq_cli/generate.py", line 278, in cli_main
    main(args)
  File "/home/attn2d/fairseq_cli/generate.py", line 36, in main
    return _main(args, sys.stdout)
  File "/home/attn2d/fairseq_cli/generate.py", line 103, in _main
    num_workers=args.num_workers,
  File "/home/attn2d/fairseq/tasks/fairseq_task.py", line 181, in get_batch_iterator
    required_batch_size_multiple=required_batch_size_multiple,
  File "/home/attn2d/fairseq/data/data_utils.py", line 220, in batch_by_size
    from fairseq.data.data_utils_fast import batch_by_size_fast
  File "fairseq/data/data_utils_fast.pyx", line 1, in init fairseq.data.data_utils_fast
    # cython: language_level=3
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject
##########

I just gave up after that.. @elbayadm I hope you can help me on this.

Code sample

Environment

I have followed the instructions in the README to install my environment. :

git clone https://github.com/elbayadm/attn2d
cd attn2d
pip install --editable .

As a result, I have the following libraries in my environment:

Python 3.7.10 | packaged by conda-forge | (default, Feb 19 2021, 16:07:37) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
torc>>> torch.__version__
'1.9.0+cu102'
>>> import fairseq
>>> fairseq.__version__
'0.9.0'
>>> 
$ python --version
Python 3.7.10

Operating system: Linux

[EricLina]

So Why the bug is not fixed for such a long time ?

[protonish]

ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject

@ereday @EricLina IIRC this error often comes up in Fairseq because of numpy version issues. If you upgrade numpy to 1.22 or something later it should be fixed (or at least it has worked for me numerous times when working on regular MT models). If you're still going down that rabbit hole, perhaps it's worth a try?